Rob🦇🔊

🇵🇪, always positive and tech lover → learning and building

[Claude @claudeai]

Effective today, we are:

1) Doubling Claude Code’s 5-hour rate limits for Pro, Max, and Team plans;

2) Removing the peak hours limit reduction on Claude Code for Pro and Max plans; and

3) Substantially raising our API rate limits for Opus models.

[☀️ @awstian]

Day 21 of Gengar. Check it out @robertointech’s article 👇

[Rob🦇🔊 @robertointech]

https://t.co/9LR3bohPqh

[Rob🦇🔊 @robertointech]

Nerfearon a Opus 4.6 🚨

[Rob🦇🔊 @robertointech]

@awstian

[☀️ @awstian]

Day 6 building Gengar OS and now 71% WR global and 92% in the predictive model.

Basically Gengar predicts the price of BTC -> 2 mins before polymarkets

Sports are not bad but with the backtest we noticed that even with good WR is not so good as we expected and is now updating with a reduced budget.

More journal sampling to catch insights + backtest and improvements.

We also have copy trading but the more time goes the less we use copy trading.

Basically copy = latency and 0 control.

Is better develop a own edge and make predictive models based in %.

[☀️ @awstian]

How to build a Polymarket bot that mathematically out-trades human whales.

Manual trading on Polymarket is a brutal game. I tracked the biggest whales and found they only average a 47.6% win rate. They are bleeding capital to fees and human bias.

So, I took my traditional strategy, added HFT mechanics, and built AutoBTC v2.

The result? A 68.9% Win Rate over 3,600+ candles.

Here is the exact alpha on how automating the micro-inefficiencies changes everything:

1. Oracle Latency Exploitation (The "Pre-Frontrun")

Humans react in seconds. My bot monitors Binance/Bitstamp in milliseconds. It captures the gap before the Chainlink oracles can update the token prices on-chain. By the time retail sees the move, we are already in.

2. Zero-Volume Market Sniping:
Retail traders are terrified of empty order books. The bot isn't. It safely snipes entries at $0.52 in markets with $0 initial volume. Why? Because Chainlink guarantees the resolution. We get our fills before the crowd even wakes up

3. Surgical Technical Exclusions
Humans suffer from FOMO. The bot strictly uses BB(20, 2.25σ) + StochRSI and requires dual timeframe confirmation (5m + 15m).

Even better: backtesting showed UTC hours 13, 19, and 23 are highly toxic (36-47% WR).

A human would force trades; the bot simply powers down. That's why sometimes I get 202 signals but take only 2 entries.

4. Dynamic Kelly Exits

Holding to a 1.00 resolution is for gamblers. The bot constantly calculates the expected value (EV). If the EV_hold drops below 0.015, the dynamic exit kicks in, securing profits and freeing up capital for better setups (Take in consideration the polymarket fee as well)

This discovery was mind-blowing and only possible through rigorous backtesting. In the past, backtesting used to be a massive pain in the ass. Now, extracting and structuring data with AI (like Claude or other tools) makes it a breeze.

You can grab 10 million historical trades plus the BTC price history, and start developing an edge in a fraction of the time. When you analyze your losers, you figure out exactly what works and what doesn't.

For example, my original edge was trading only during specific hours. But even then, I noticed I had many trades sitting at 30% or 50% in profit that eventually got wrecked by mean reversion. Now, the dynamic EV filter catches those exits perfectly. Most bots only use a static EV...this dynamic adjustment is a massive game-changer.

By the way, I'm still tunning my bot but the WR now is impressive, still recording samples for my journal and fine tunning it.

So, is this all you need to build a profitable bot? Hell naaa...

But these 4 points won't just upgrade your bot, they will completely rewire how you think about algorithmic architecture.

I know guys building bots who are blown away by having access to such a "massive" data package. Meanwhile, I'm already figuring out how to scrape even more granular, specialized data. Why? Because you can't improve what you don't measure.

If your metrics are dialed in, this system prints, not just because of the strategy itself, but because of the sheer speed advantage I mentioned earlier.

There are plenty more cherries on top of this cake, but to get that alpha, you'll have to drop a follow👻

[Google Research @GoogleResearch]

Introducing TurboQuant: Our new compression algorithm that reduces LLM key-value cache memory by at least 6x and delivers up to 8x speedup, all with zero accuracy loss, redefining AI efficiency. Read the blog to learn how it achieves these results: https://t.co/CDSQ8HpZoc

[Claude @claudeai]

New in Claude Code: auto mode.

Instead of approving every file write and bash command, or skipping permissions entirely, auto mode lets Claude make permission decisions on your behalf.

Safeguards check each action before it runs.

[Rob🦇🔊 @robertointech]

Me encanta la tranquilidad con lo que lo anuncian jajajaa

[Claude @claudeai]

You can now enable Claude to use your computer to complete tasks.

It opens your apps, navigates your browser, fills in spreadsheets—anything you'd do sitting at your desk.

Research preview in Claude Cowork and Claude Code, macOS only.

[bunny @ConejoCapital]

im raising a $1B VC fund with the only thesis to invests in founders that look like this. DM with serious inquiries only

[Rob🦇🔊 @robertointech]

@awstian @ConejoCapital Yeah!!!

[☀️ @awstian]

@ConejoCapital @robertointech and I are building something

[Claude @claudeai]

New in Cowork: scheduled tasks.

Claude can now complete recurring tasks at specific times automatically: a morning brief, weekly spreadsheet updates, Friday team presentations.

[Rob🦇🔊 @robertointech]

@codeup201918017 @EthereumLima Asiste el sábado al meeting presencial

[Rob🦇🔊 @robertointech]

After a week to Dis-Connect from DevConnect, I’m finally back — fully recharged from the buzz.
It’s always a pleasure to work side-by-side with the @devcon team and share incredible moments with fellow volunteers. Swag Station is the most requested booth for a reason… and I’m thrilled I got to be part of it again. We do swag with actual swag 😎 iykyk.

Now that the chaos has settled, I’m ready for the next challenge.
If you’re looking for someone who lives and breathes operations, stays positive under pressure, and has a heart fully Ethereum-aligned — I’m your person.

From EA to Operations Manager, I get things done and I go beyond.
These past years I’ve been all-in on education, open source, and public goods through @ethkipu @giveth @ethereumlima @cryptoversidad and our little dev community @cevidevs.

Here to keep making Ethereum for everyone.
My DMs are open 💛

[Rob🦇🔊 @robertointech]

@incakolitazero @EFDevcon

[Rob🦇🔊 @robertointech]

first vs last @EFDevcon

[Rob🦇🔊 @robertointech]

@youfoundperrie @EFDevcon

[Rob🦇🔊 @robertointech]

@lain_calvo @UNRoficial Qué grande!

[Rob🦇🔊 @robertointech]

@OrnellaWeb3 @EFDevcon